Upload
others
View
17
Download
0
Embed Size (px)
Citation preview
JOURNAL OF CHEMICAL PHYSICS VOLUME 112, NUMBER 19 15 MAY 2000
Synchronous helical pulse sequences in magic-angle spinning nuclearmagnetic resonance: Double quantum recoupling of multiple-spin systems
Andreas Brinkmann, Mattias Eden, and Malcolm H. Levitta)
Division of Physical Chemistry, Arrhenius Laboratory, Stockholm University, S-106 91 Stockholm, Sweden
~Received 6 December 1999; accepted 19 January 2000!
Some general principles of radio-frequency pulse sequence design in magic-angle spinning nuclearmagnetic resonance are discussed. Sequences with favorable dipolar recoupling properties may bedesigned using synchronous helical modulations of the space and spin parts of the spin Hamiltonian.The selection rules for the average Hamiltonian may be written in terms of three symmetry numbers,two defining the winding numbers of the space and spin helices, and one indicating the number ofphase rotation steps in the radio-frequency modulation. A diagrammatic technique is used tovisualize the space-spin symmetry selection. A pulse sequence C144
5 is designed whichaccomplishes double-quantum recoupling using a low ratio of radio frequency field to spinningfrequency. The pulse sequence uses 14 radio frequency modulation steps with space and spinwinding numbers of 4 and 5, respectively. The pulse sequence is applied to the double-quantumspectroscopy of13C3-labeled L-alanine. Good agreement is obtained between the experimental peakintensities, analytical results, and numerically exact simulations based on the known moleculargeometry. The general symmetry properties of double quantum peaks in recoupled multiple-spinsystems are discussed. A supercycle scheme which compensates homonuclear recoupling sequencesfor chemical shifts is introduced. We show an experimental double-quantum13C spectrum of@U–13C#–L-tyrosine at a spinning frequency of 20.000 kHz. ©2000 American Institute ofPhysics.@S0021-9606~00!01214-9#
piio
s
.le
yi
spinoo
naig
o
gth
-
er
s ofheytedil-in
ofto
they toater-or-
de-lerdered
le-ldsies
-r
fre-
o-
I. INTRODUCTION
Solid state nuclear magnetic resonance~NMR! may beused for determining local molecular structures in isotocally labeled biomolecules. Methods exist for determinatof accurate intermolecular distances,1–17 and molecular tor-sional angles.18–24 The methods have been applied to sytems such as noncrystalline membrane proteins25 providinginformation which is currently inaccessible any other way
Most realistic applications require rapid magic-angspinning~MAS! in order to obtain maximal signal intensitand spectral resolution. Molecular structural informationobtained by applying radio frequency~rf! pulse sequencewhich implement temporary recoupling of the nuclear ssystem, in order to take advantage of the dipole–dipole cplings between the nuclear spins. The radio frequency recpling schemes may be incorporated into multidimensioprocedures, leading to powerful methods for spectral assment and geometry determination.15,26–29
Most of the existing recoupling methods fall into twbroad classes, depending on whether they generate~in idealcircumstances! a zero-quantum or double-quantum averaHamiltonian for the recoupled spin pairs. For example,rotational resonance,1–6 RFDR,26 and RIL12 schemes arezero-quantum recoupling methods. The HORROR13 andC714 schemes and their variants,15–17are double-quantum recoupling techniques. The DRAMA9 and DRAWS10 se-quences produce mixtures of double-quantum and z
a!Electronic mail: [email protected]
8530021-9606/2000/112(19)/8539/16/$17.00
-n
-
s
u-u-l
n-
ee
o-
quantum average Hamiltonians. These alternative moderecoupling may be complementary, in the sense that toperate best in different circumstances. It has been poinout12 that evolution under a zero-quantum average Hamtonian conserves the total longitudinal magnetizationmultiple-spin systems, leading to a uniform distributionmagnetization at long mixing intervals. This correspondsthe solution-state TOCSY method,30 which is very useful forassignment purposes. The double-quantum methods, onother hand, allow the use of double-quantum spectroscopestablish assignments, as in the solution-stINADEQUATE experiment,31–33 at the same time as ensuing complete suppression of isolated spin signals, an imptant advantage in large biomolecules.
In addition, double-quantum techniques have beenveloped which are insensitive to one of the three Euangles determining the orientation of molecules in a powsample. This leads to high efficiency in nonorientsamples.13,14So far only rotational resonance1–6 achieves thisin the case of zero-quantum recoupling.
One of the technical problems affecting current doubquantum recoupling sequences is the need for high rf fieat the Larmor frequencies of the recoupled spin spec~henceforth denoted byS!. For example, the C7,14
POST-C7,16 and CMR715 methods require theS-spin nuta-tion frequencyvnut
S to be 7 times the sample spinning frequency,vnut
S 57v r . This is not usually a severe problem fosamples containing only one type of spin, since nutationquencies of around 120 kHz are routinely available.27 How-ever, the situation is more difficult in the context of biom
9 © 2000 American Institute of Physics
fo-ehealthline
le
igh
cthe
if
mrymsuio
a
nil
road
-th
efots
en
d.es
-mr
himghrm
cuolth
ronuw
le
ingallgate
de-aksctdi-o-
re-icalofthe
eri-
rf
s
o-iss
herota-f.ase-in
t ofas-on
ac-ith
e-s
me
8540 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
lecular NMR, where theS-spins usually take the form ointroduced13C labels immersed in a pool of abundant prtons ~henceforth denotedI!. It is necessary to decouple thI-spins from theS-spins at the same time as recoupling tS-spins with each other. Heteronuclear decoupling is usuachieved by applying a strong unmodulated rf field atI-spin Larmor frequency, at the same time as the recouprf field on theS-spins. It has been found empirically that thI-spin nutation frequency must approach 3 times theS-spinnutation frequency in order to achieve good heteronucrecoupling in this context,vnut
I *3vnutS .15 This is a restrictive
condition, since most NMR probes do not tolerate very hrf fields on two irradiation channels at the same time. Tneed for good heteronuclear recoupling has so far restrithe biomolecular applications of C7 and its relatives to ratlow spinning frequencies~typically around 6 kHz!, whichleads to a loss of sensitivity, mainly due to chemical shanisotropy modulations during the signal acquisition.34
The rf field requirements of C7 may be reduced by eploying fivefold symmetry instead of sevenfold symmetThis introduces additional error terms which must be copensated by doubling the length of the sequence. The reing cycle is called SPC-5 and has the matching conditvnut
S 55v r .17
Recently, we showed that the symmetry arguments leing to the original C7 sequence may be generalized.35 Aclass of pulse sequences denoted CNn
n has been described, iwhich the symmetry-allowed terms in the average Hamtonian are chosen according to simple theorems. The plem of heteronuclear decoupling in rotating solids wasdressed using the CNn
n concept.35
In the current paper, the nature of CNnn sequences is
explored further. The CNnn symmetry is identified as provid
ing synchronized helical modulations of the space andspin parts of the spin Hamiltonian. The labelsm andn indi-cate thewinding numbersof the space and spin helices. Wshow that generalized helical symmetry admits solutionshomonuclear recoupling with lower rf field requirementhan the original C7 sequences. We describe a new sequdenoted C144
5, which has an rf field requirement ofvnutS
53.5v r , i.e., half of that required in the original C7 methoThis allows application at fairly high spinning frequenciwhile still achieving good heteronuclear decoupling.
The pulse sequence C1445 permits operation at high spin
ning frequencies but is not very well compensated for checal shift perturbations. In order to further improve its perfomance we employ a supercycle scheme which combinesdouble-quantum efficiency with the compensation for checal shifts. The supercycled sequence, called SC14, is rouas broadband as previous recoupling schemes, while peting application at higher spinning frequency.
Double-quantum recoupling is expected to be partilarly useful in the spectroscopy of heavily labeled biomecules, which are often easier and cheaper to synthesizeselectively labeled substances. Multidimensional spectcopy will play an important role in spectral assignment, ain deriving multiple geometric constraints by simultaneodistance and angle estimations. Success of this approachrequire a solid understanding of the dynamics of recoup
lyeg
ar
heedr
t
-.-lt-n
d-
-b--
e
r
ce,
i--ghi-lyit-
--an
s-dsilld
multiple-spin systems. In general, the effective recouplHamiltonian induces global spin dynamics which involverecoupled spins at the same time. In this paper we investithe dynamics of double-quantum excitation in a13C3 cluster.We obtain quantitative agreement between the amplitudevelopment of experimental two-dimensional spectral pewith analytical results, as well as with numerically exasimulations using the known molecular geometry. In adtion we explore the theoretical phase properties of twdimensional double-quantum spectra in multiple-spincoupled systems and present approximate analytexpressions for individual peak amplitudes, as a functionthe excitation and reconversion intervals. To demonstratepractical utility of the new sequences, we show an expmental double-quantum spectrum of@U–13C#–L-tyrosine,obtained at a spinning frequency of 20.000 kHz.
II. HELICAL RECOUPLING SEQUENCES
A. Synchronous helical modulations
The symbol CNnn refers to a set of rotor synchronized
pulse cycles, with the following properties:~i! Each rf cyclehas a durationtC5nt r /N, wheret r5u2p/v r u is the rotationperiod, andv r is the sample rotation frequency. This impliethat N rf cycles are timed to coincide withn sample rotationperiods.~ii ! Each rf cycle is designed to provide no net evlution of the nuclear spin states, when only the rf fieldtaken into account.~iii ! The rf phase of consecutive cyclediffers by 2pn/N. The phase of thepth cycle is thereforegiven byFp5F012pnp/N, with p50,1,2,... . HereF0 isthe initial phase of the whole block.
The duration of an entire CNnn sequence is denotedT
5NtC. The symmetry numbersn andn may be thought ofas ‘‘winding numbers’’ of two helices, one representing tspatial sample rotation, and one representing the phasetions of the rf fields. A complete sequence consists oncomplete sample rotations andn complete rf phase rotationsThe sample rotation is continuous, while the pulse phrotations are performed inN discrete steps. A pictorial representation of this concept is given for two different casesFig. 1.
In the general case, each element C may itself consispulses with different phase. However, in this paper, wesume that the rf field is only subject to amplitude modulati~or a p phase shift! within each cycle C.
B. Space-spin selection rules
Consider a set ofN S-spins, denotedS1 ,S2 ,...,SN , ex-periencing spin–spin interactions and chemical shift intertions as well as a rotor-synchronized rf pulse sequence wthe symmetry CNn
n . The spin interaction terms are convniently described in the interaction frame of the rf field. Adescribed in Refs. 14, 15, and 35, the interaction fraHamiltonian at time pointt may be written
H~ t !5 (L,l ,m,l,m
H lmlmL ~ t !, ~1!
n-ctpa
u-
s
ou
Exin
et
a
-
s.ir-ledla-
ndetry
sor
the
the
e tone
the
intoll asown
sota-s
if-f tf
8541J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Helical pulse sequences in MAS–NMR
where the symbolL represents the type of interactio~chemical shift, spin–spin coupling!, and the quantum numbersl, m, l, m index the symmetry of the term with respeto rotations of the spin polarizations and with respect to stial rotations of the sample. The termH lmlm
L (t) transforms asan irreducible spherical tensor of rankl for spatial rotationsand rankl for spin rotations. The component indicesm andm have valuesm52 l ,2 l 11,...,l for space andm52l,2l11,...,l for spin. For example, the direct dipole–dipole copling between homonuclear spin pairs has ranksl 52, l52;the J-coupling between homonuclear spin pairs has rankl50, l50; the isotropic chemical shift has ranksl 50, l51;the chemical shift anisotropy and heteronuclear dipolar cplings have ranksl 52, l51. All components withl 52, m50 vanish in the case of exact magic-angle spinning.plicit expressions for the various terms may be foundRefs. 14 and 15.
As shown in Refs. 14, 15, and 35, the CNnn symmetry of
the pulse sequence imposes the following periodic symmon the interaction frame terms:
H lmlmL ~ t1ptC!5H lmlm
L ~ t !expH i2p~mn2mn!p
N J . ~2!
The results of the sequence may be analyzed using the Mnus expansion:36,37
H~ t !5H ~1!1H ~2!1H ~3!1¯, ~3!
where the first two orders36 are given by
H ~1!5T21Et0
t01Tdt H~ t !, ~4!
FIG. 1. Visualization of the synchronized helical modulations for two dferent CNn
n sequences. In each case the rotation of the space part oHamiltonian~sample rotation! is shown by a continuous spiral trajectory othe rotor phase, which completesn full revolutions. The trajectory of thespin part ~rf phase! completesn full revolutions in the same time, inNdiscrete steps.~a! Visualization of C72
1 symmetry.~b! Visualization of C1445
symmetry.
-
-
-
ry
g-
H ~2!5~2iT !21Et0
t01Tdt8E
t0
t8dt@H~ t8!,H~ t !#, ~5!
if the sequence starts at time pointt0. The first order resultfor the effective Hamiltonian is
H ~1!5 (L,l ,m,l,m
H lmlmL~1! , ~6!
where
H lmlmL~1! 5T21E
t0
t01Tdt Hlmlm
L ~ t !. ~7!
As shown in Ref. 35, Eq.~2! leads to the following symmetry theorem:
H lmlmL~1! 50 if mn2mnÞN3 integer. ~8!
Similar theorems exist for the higher order Magnus term35
The result Eq.~8! allows the design of sequences with desable recoupling properties without calculation of the detaistructure of the cycle C, at least in the first stage of calcution.
The magnitude of the symmetry allowed terms depeon the details of the pulse sequence. In general, a symmallowed term has the form
H lmlmL~1! 5k lmlm@Alm
L #R exp$2 im~aRL0 2v r t
0!%TlmL , ~9!
where@AlmL #R is a space component of the interaction ten
L, written in the rotor-fixed frame, andaRL0 denotes the ini-
tial rotor position. For sequences involving onlyp phaseshifts, the scaling factork lmlm of a symmetry allowed termwith the quantum numbers (l ,m,l,m) is given by
k lmlm5 i mdm0l ~bRL!tC
21
3E0
tCdt dm0
l ~2bnut~ t !!exp$ imv r t%, ~10!
wherebRL defines the angle between the rotor axis andfield, and the rf nutation angle is
bnut~ t !5Et0
t01tdt8 vnut~ t8! ~11!
as described more fully in Ref. 35.The consequences of Eq.~8! are conveniently explored
using space-spin selection diagrams~SSS diagrams!38 asshown in Figs. 2 and 3. These diagrams are similar tocoherence transfer pathway diagrams~CTP diagrams!39 usedin the design of phase cycling schemes. The resemblancCTP diagrams is not coincidental: CTP diagrams help ovisualize symmetry selection of terms under rotations ofspin polarizations around the field axis~coherence order!. Ina rotating sample exposed to an rf field, one must takeaccount the effect of macroscopic sample rotation, as wespin rotations, and this is the task of the SSS diagrams shin the current paper.
Figure 2 shows SSS diagrams for the C721 sequence. The
plotted levels indicate the total value ofmn2mn. The su-perposition ofmn and2mn is broken into two stages, so ato separate out the effects of spatial rotations and spin rtions. The ‘‘barrier’’ at the right-hand side of the diagram
he
e
ome
ore-tue-, oth
th.
mpleres,o
n--icit
b
spe
n-
ex-for
the3.t of
hepli-ich
d
ger
e
r
8542 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
has holes separated byN units, which corresponds to thinequality in the symmetry theorem Eq.~8!. Pathways whichpass through a hole in the barrier indicate space-spin cponents which are symmetry allowed in the first order avage Hamiltonian.
Figure 2~a! shows that all CSA components~m5$61,62% andm5$0,61%! are suppressed by C72
1 symmetry in thefirst order average Hamiltonian. Figure 2~b! shows that onlyhomonuclear dipolar components with (m,m)5(1,2) aresymmetry allowed@and by implication, also (m,m)5(21,22)#. The selection of terms withm562 indicates double-quantum recoupling of the nuclear spin system. Furthermthe fact that them512 term is associated with only onspatial rotational component (m51) is associated with a favorable orientation dependence of the double-quanexcitation.14 One of the orientational Euler angles is ‘‘phasencoded,’’ implying that the phase, but not the amplitudethe recoupled double-quantum Hamiltonian depends onvalue of this angle. This property is partly responsible forhigh efficiency of C7 sequences in nonoriented samples14
The symmetry properties of C721, as visualized in Fig. 2,
are independent of the detailed structure of C. The optichoice of C is dictated by other considerations, for examthe magnitude of the symmetry-allowed terms, the suppsion of interference from isotropic chemical shiftand the robustness of the sequence with respect tinhomogeneity. The cycles C05(2p)0(2p)p and C0
5(p/2)0(2p)p(3p/2)0 have both been exploited in the cotext of C72
1 symmetry.14,16 Both cycles are internally compensated for rf inhomogeneity effects, and the latter chogives an overall sequence which is particularly robust wrespect to chemical shifts and rf inhomogeneity.16
The solutionN57, n52, n51 is far from unique. TableI shows some additional solutions for phase encoded douquantum recoupling. The symmetry C75
1 has been pointedout before.15,38 The symmetries C71
3, C914, C111
5, and C1316
are particularly interesting, since they permit selective phaencoded double-quantum recoupling within a single rotorriod.
In addition, the interaction frame symmetry Eq.~2! maybe generated by pulse sequences in which the rf fields doobviously conform to the CNn
n pattern. This has been dis
FIG. 2. Space-spin selection diagram~SSS diagram! for C721. ~a! Suppres-
sion of all CSA modulation components.~b! Selection of a single 2Qdipole–dipole component, with quantum numbers (m,m)5(1,2). The mir-ror image pathways stemming fromm521, m522 have been suppressefor simplicity.
-r-
e,
m
fe
e
al,
s-
rf
eh
le-
e--
ot
cussed briefly in the context of TPPM decoupling.35 In aforthcoming paper, we show that this concept may beploited to generate even more restrictive selection rulesthe average Hamiltonian.40
C. rf amplitude compensation
In this paper, we concentrate on the properties ofsequence C144
5, which has the SSS diagram shown in Fig.The space-spin selection properties are identical to thaC72
1, except that the components (m,m)5(21,2) and~1,22! are selected instead of~1, 2! and ~21, 22!.
Although the SSS properties of C721 and C144
5 are essen-tially equivalent in the first order average Hamiltonian, tsequences differ in their robustness with respect to rf amtude errors. The reason for this is illustrated in Fig. 4, wh
FIG. 3. Space-spin selection diagram~SSS diagram! for C1445. ~a! Suppres-
sion of all CSA modulation components.~b! Selection of a single 2Qdipole–dipole component, with quantum numbers (m,m)5(1,22). Thevertical scale of this figure is one half that of Fig. 2. The mirror imapathways stemming fromm521, m522 have been suppressed fosimplicity.
TABLE I. A list of inequivalent CNnn symmetries withN<14, n<5 and
0<n<N/2, suitable for g-encoded double-quantum recoupling. Thsymmetry-allowed terms withm561 and m562 are shown. All otherterms withm561,62 andm50,61,62 are suppressed in the first ordeaverage Hamiltonian. Additional variants with similar properties for CNn
n
are given by CNnZN6n whereZ is an integer.
Sequence (m,m) Sequence (m,m)
C713 ~21,2! ~1,22! C73
2 ~21,2! ~1,22!C91
4 ~21,2! ~1,22! C1134 ~21,2! ~1,22!
C1115
C1316
~21,2!~21,2!
~1,22!~1,22!
C1335 ~21,2! ~1,22!
C742 ~21,22! ~1,2!
C721 ~21,22! ~1,2! C94
2 ~21,22! ~1,2!C82
1 ~21,22! ~1,2! C1043 ~21,2! ~1,2!
C823 ~21,2! ~1,22! C114
2 ~21,22! ~1,2!C92
1 ~21,22! ~1,2! C1342 ~21,22! ~1,2!
C1021
C1121
~21,22!~21,22!
~1,2!~1,2!
C1445 ~21,2! ~1,22!
C1221 ~21,22! ~1,2! C75
1 ~21,2! ~1,22!C122
5 ~21,2! ~1,22!C132
1 ~21,22! ~1,2! C952 ~21,2! ~1,22!
C1153 ~21,2! ~1,22!
C1354 ~21,2! ~1,22!
nced
letl
u-le
ii
-
e
s
atto
rssCa7
thitethig
ncin
totha
haal-
ofm-
ed
the
nef
ec
tyl
am-
in-
ncy
oss-am-xci-
Hz
ct
8543J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Helical pulse sequences in MAS–NMR
shows the sequence of rf phases used in the two sequeIn Fig. 4~a! the sequence of rf fields in the transverse planillustrated for the case of C72
1. The phases are incrementebetween cycles in steps of 2p/7. Suppose that each cycrotates the spin polarizations through an angle slighgreater than 2p. In a first order approximation, the accumlated rotations cancel out after a full set of seven cycHowever, this cancellation is established very slowly, andpractice, only very small rotation errors are tolerated. Ittherefore necessary to use the symmetry C72
1 in combinationwith a cycle which isinternally compensated for rf amplitude errors. The cycles C05(2p)0(2p)p and C0
5(p/2)0(2p)p(3p/2)0 fulfill this condition, at the expenseof requiring an rf field which is twice as large as for thsimplest possible sequence C05(2p)0 , which is uncompen-sated for rf field amplitude errors.
The situation for C1445 is different, as shown in Fig. 4~b!.
Here the step in phase between consecutive cycle10p/14>129°. Since this is close to the angle 2p/35120°,each group of three consecutive cycles is well compensfor rotation errors. This may be visualized by addinggether the bold vectors in Fig. 4~b!. The symmetry C144
5 istherefore intrinsically compensated for rf amplitude erroand internal compensation of the cycle C may be dispenwith. It is feasible to use the simplest possible cycle05(2p)0 , allowing a reduction of the required rf field byfactor of 2, for a given spinning frequency, compared to C2
1
sequences employing cycles of the form C05(2p)0(2p)p
or C05(p/2)0(2p)p(3p/2)0 .This compensation mechanism is independent of
sign of n. However, in practice sequences with oppossigns ofn have slightly different performances, becausechemical shift anisotropy interactions have a defined sThis is illustrated by the simulations in Fig. 5~a!, which showthe calculated double-quantum filtered efficiency as a fution of rf irradiation frequency, for parameters correspondto @13C2,15N#-glycine at a field of 9.4 T.41 In most of theexperimental results shown below, we employ C144
25 insteadof C144
5. In all experimental implementations we take inaccount the sense of the Larmor frequency as well asradio frequency mixing scheme, as described in Refs. 4243.
Experimental evidence of the rf compensation mecnism of C144
25 is shown in Fig. 6, which shows experimentresults for @13C2,15N#-glycine, obtained at a spinning fre
FIG. 4. ~a! rf fields used in C721, denoted as vectors in the transverse pla
The numbers refer to the cycle indexq50,1,...,6. The first three elements othe sequence are denoted as bold vectors.~b! rf fields used in C144
5. Thenumbers refer to the cycle indexq50,1,...,13. The first three elements of thsequence are denoted as bold vectors. Consecutive phase values whichbine to give good rf error compensation are drawn with the same line s
es.is
y
s.ns
is
ed-
,ed
e
en.
-g
end
-
quency ofv r /2p511.000 kHz. The figure shows integralsdouble-quantum-filtered spectra as a function of rf field aplitude for the sequences C72
1 and C14425. Both sequences
use an uncompensated elemental cycle C05(2p)0 . The C721
symmetry performs rather poorly in this case. The improvrobustness of the C144
25 cycle is visually obvious.It should be noted that the sequence C144
25 may be de-rived from two consecutive C72
1 sequences, by addingp tothe phase of odd numbered C elements, and continuingpattern through four rotor periods.
.
om-e.
FIG. 5. Simulated double-quantum filtered efficiencies for~a! SC14, C14425,
C1445; ~b! SC14, POST-C7, SPC-5. The simulations are done for the par
eters of@13C2#-glycine at a field ofB059.4 T ~Ref. 41!. Powder averagingwas performed using 538 orientationsVMR , chosen according to the ZCWscheme ~Ref. 62!. In all cases the rf amplitude is given byvnut
S /2p570.070 kHz. The spinning frequency, the excitation and reconversiontervals are given by SC14, C144
5, C14425: v r /2p520.020 kHz, tE5tR
5599.4ms; POST-C7:v r /2p510.010 kHz, tE5tR5599.4ms; SPC-5:v r /2p514.014 kHz,tE5tR5713.6ms. The theoretical maximum of 73%is indicated by a solid line.
FIG. 6. Experimental measurements of double-quantum filtered efficiefor C72
1 and C14425 sequences, obtained on@13C2,15N#-glycine ~98% 13C,
96–99 %15N!, at a field ofB054.7 T and a spinning frequency ofv r /2p511.000 kHz using a Chemagnetics-Infinity-200 spectrometer, and a crpolarization contact time of 1.4 ms. The sample was purchased from Cbridge Isotope Laboratories and used without further purification. The etation and reconversion intervals were bothtE5tR5519.4ms. Continuouswave decoupling was applied with the proton nutation frequency 122 kduring CNn
n , and 86 kHz during acquisition. An elemental cycle C0
5(2p)0 was used in both cases. The13C nutation frequency is varied alongthe horizontal axis. The C144
25 sequence is clearly more robust with respeto rf amplitude errors.
so
lehre
-le
nra
e
ies.r-ift
r-ce
8544 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
As shown in Table I, there are many other feasiblelutions for the symmetry numbersN, m, andn. For example,the symmetry C75
1 may be used in conjunction with the cycC05(2p)0(2p)2p/3(2p)4p/3 to provide a sequence witgood rf amplitude compensation and low rf power requiments:vnut
S 54.2v r .38 The cycles C05(2p)0(2p)p and C0
5(p/2)0(2p)p(3p/2)0 are poor choices in this context, because they lead to very small magnitudes for the recoup2Q term.15 The cycle C05(2p)0(2p)2p/3(2p)4p/3 does notsuffer from this defect and leads to a C75
1 sequence which isalmost as efficient as the C144
25 sequence based on C0
5(2p)0 . Table I presumably contains many other solutiowith favorable properties, some of which may allow opetion at even higher spinning frequency than C144
25.
D. Supercycle construction
The sequence C14425 with C05(2p)0 has fairly good rf
error compensation but is not well compensated with resp
Tti.
of
orC
atth
afo
14le-rthlinerd
n-ind
-
-
d
s-
ct
to resonance offset effects and chemical shift anisotropThis is illustrated in Fig. 5~a!. We have constructed supecycles of C144
25 which have an acceptable chemical shcompensation. One example is the supercycle
SC145C1445•@P0
21•C144
25•P0#p/7•@C144
5#p
•@P021•C144
25•P0] 8p/7 ,
where the notationP0 indicates the insertion of ap pulseelement with phasef50 andP0
21 indicates the deletion of ap-pulse element. The notation@¯#f indicates an overallphase shift of the bracketed sequence byf. The sequence@P0
21•C144
25•P0#p/7 is therefore a phase-shifted cyclic pe
mutation of the C14425 sequence. The full SC14 sequen
may be written as follows:
3600 360128.57 360257.14 36025.71 360154.29 360282.86 36051.43
360180 360308.57 36077.14 360205.71 360334.29 360102.86 360231.43
18025.71 360257.14 360128.57 3600 360231.43 360102.86 360334.29
360205.71 36077.14 360308.57 360180 36051.43 360282.86 360154.29 18025.71
360180 360308.57 36077.14 360205.71 360334.29 360102.86 360231.43
3600 360128.57 360257.14 36025.71 360154.29 360282.86 36051.43
180205.71 36077.14 360308.57 360180 36051.43 360282.86 360154.29
36025.71 360257.14 360128.57 3600 360231.43 360102.86 360334.29 180205.71
~12!
ed.r
e
-
e-ts
ts at
re-em-
where all flip angles and phases are specified in degrees.complete sequence spans 16 rotor periods. The theoreprinciples of this supercycle are discussed in Appendix A
The performance of SC14 with respect to resonanceset is illustrated in Fig. 5~solid lines!. Figure 5~b! shows thatthe maximum double-quantum efficiency and offset perfmance are comparable to those of POST-C7 and SP~dashed and dotted lines! under these conditions. Note thall simulations are performed at the same rf field strengcorresponding to a nutation frequency ofvnut
S /2p570.070 kHz. However, the sample spinning frequenciesdifferent in the three cases, corresponding to 10.010 kHzPOST-C7, 14.014 kHz for SPC-5, and 20.020 kHz for SCFigure 5~b! illustrates the point that SC14 permits doubquantum excitation at higher spinning frequencies than pvious pulse sequences, in the case that the rf field strenglimited by probe performance and heteronuclear decouprequirements. In multiply labeled spin systems, it is genally desirable to rotate the sample as fast as possible, in oto achieve optimal sensitivity and resolution.44
Although SC14 performs well at high spinning frequecies, simulations indicate that it performs poorly at low spning frequencies. Sequences such as POST-C7 shoulused in this regime.
hecal
f-
--5
,
rer.
e-isg
r-er
-be
III. DOUBLE-QUANTUM SPECTROSCOPY OFMULTIPLE-SPIN SYSTEMS
A. Pulse sequence
Double-quantum spectroscopy of13C-labeled organicsolids at high MAS spinning frequencies may be performusing the radio frequency~rf! pulse sequence shown in Fig7. The row markedI denotes the rf fields at the Larmofrequency of the abundant protons, whileS denotes the rffields applied at the13C Larmor frequency. The sequencstarts with ramped cross polarization to enhance theS-spinmagnetization.45 The followingp/2-pulse converts the transverse magnetization into longitudinalS-spin magnetization.The ramped cross-polarization field and thep/2 pulse havethe phasesFprep andFprep2p/2 respectively, whereFprep isthe overall rf phase of the preparation interval.
The p/2-pulse is followed by a 2Q-excitation pulse squence of durationtE . The 2Q-excitation sequence convertheS-spin longitudinal magnetization into~62!-quantum co-herence. The double-quantum excitation sequence starthe time pointtE
0 and terminates at time pointtE15tE
01tE .In the following discussion, we assume that a C144
25
sequence is used for the double-quantum excitation andconversion as shown in Fig. 7. The SC14 supercycle is donstrated later on in this paper.
di-
e
en
ta
tein
fofu-
hi
tase
eme
eral
rmmn-
ed in
for
Iningo-
e is
e rf
-
risk
tro
ig
8545J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Helical pulse sequences in MAS–NMR
The excitation part of the sequence consists ofqE cycles,corresponding to an excitation interval oftE5qEtC. Theoverall phase of the excitation block denoted isFE so thatthe rf cycles have phasesFE , FE210p/14, FE220p/14,etc. Note that the number of cyclesqE is not restricted to bea multiple of 14.
The excited double-quantum coherences are alloweevolve for an intervalt1 , and are reconverted into longitudnal S-spin magnetization by applying anotherqR cycles ofC144
25 irradiation. The reconversion block has durationtR
5qRtC and an overall phaseFR , so that the rf phases argiven byFR , FR210p/14, FR220p/14, etc. The longitu-dinal magnetization created by the second C144
25 sequence isconverted into observable magnetization by ap/2 read pulse,whose phase is denotedF read.
The complex NMR signal is detected in the subsequperiod using a rf receiver phaseF rec and post-digitizationphase shiftFdig .42 A two-dimensional data matrixs(t1 ,t2)is compiled by acquiring a set of transients with incremention of the intervalt1 .
The specification of the rf phases is quite complicabecause of~i! phase cycling in order to select signals passthrough ~62!-quantum coherence in thet1 interval; ~ii ! thespecial phase-timing relationships which are requiredCNn
n sequences in the case that an integer number ofcycles is not completed;~iii ! the time-proportional phase incrementation~TPPI! procedure for separating the~62!-quantum signals.39
The phasesFprep, FE , FR , F read, F rec, andFdig areconveniently specified in terms of the transient counterm2
and the evolution increment counterm1 . The counterm2
50,1,...,15 is incremented on every acquired transient, wm1 is incremented between different values oft1 . The phasespecifications are
Fprep5FE5p
4m1 ,
FR~ t1!5FR0~ t1!1
p
2m2 ,
FIG. 7. Radio frequency pulse sequence for double-quantum 2D speccopy using C144
25. The phasesFprep, FE , FR , F read refer to overall rfphases of the pulse sequence blocks. The rf receiver phase during sdetection is denotedF rec and the post-digitization phase byFdig .
to
t
-
dg
rll
le
F read5p
21
p
2m21
p
2floorS m2
4 D , ~13!
F rec50,
Fdig52Frec22~FR~ t1!2FR0~ t1!!1F read2
p
2,
where the function floor~x! returns the largest integer nogreater thanx. The base reconversion phase for the phFR
0(t1) is given by
FR0~t1!5
p
21
1
2v r~ tR
02tE0 !, ~14!
wheretE0 and tR
0 are the time points marking the start of thtwo C144
25 sequences. The interval between these tipoints depends on the evolution intervalt1 according to
tR02tE
05tE1t1 ~15!
as may be seen in Fig. 7. The link Eq.~14! between the pulsesequence phases and timings allows a completely genincrementation of the evolution interval.41,46 A more re-stricted version of this result was given earlier by Hong.47 Inpractice a more specific form of Eq.~14! has proven to beuseful, which for C144
25 is given by46
FR0~ t1!5
p
22
10p
4qE1
1
2v r t1 . ~16!
The data matrixs(t1 ,t2) is subjected to a complex Fourietransform in thet2 dimension, and a cosine Fourier transforin the t1 dimension, in order to obtain the 2D spectruS(v1 ,v2). Under suitable conditions, the 2D spectrum cotains pure absorption double-quantum peaks as discussthe next section.
B. Average Hamiltonian and pulse sequencepropagators
Although C7-like pulse sequences have been usedtwo-dimensional double-quantum spectroscopy,27,47 an ex-plicit theory of this experiment has not been given so far.the following sections, we develop this theory, concentratparticularly on the amplitudes and phases of the twdimensional peaks.
Suppose that a double-quantum recoupling sequencapplied to a set ofN coupledS-spins, denotedS1 ,S2 ,...,SN .The sequence starts at time pointt0, with an overall rf phaseF0. The time-independent average Hamiltonian under thpulse sequence is given by
H ~1!5(j ,k
H jk~1! , ~17!
where the sum is taken over all spin pairs. For C721, C144
25
and many other double-quantum recoupling sequences,13,15,17
the average HamiltonianH jk(1) for a single molecular orienta
tion has the following form:
H jk~1!5v jk*
12Sj
1Sk11v jk
12Sj
2Sk212pJjkSj•Sk , ~18!
whereJjk is theJ-coupling between spinsSj andSk , v jk therecoupled through-space dipolar interaction and the aste
s-
nal
on
hle
ou
or
lin
-
rEq
l r
u-
e
ul
or
o-
ethe
ex-n-
nt
nce
er
al
alu-it-
alcts:
e
8546 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
denotes the complex conjugate. For the specific caseC144
25, the recoupled double-quantum dipolar interactidepends on the molecular orientation, the time pointt0, andthe rf phaseF0 according to
v jk~VMR ,t0,F0!5A6bjkkei ~2vr t01aRL
01gMR12F0!
3 (m522
2
d0m~2!~bPM
jk !dm21~2! ~bMR!
3e2 im~gPMjk
1aMR!, ~19!
wherednm(2) is a reduced Wigner element48 andk corresponds
to k221222 in Eq. ~10!. The Euler angles VPMjk
5$aPMjk ,bPM
jk ,gPMjk % describe the transformation of eac
dipole–dipole coupling from its principal axis to a molecufixed frame. The Euler anglesVMR5$aMR ,bMR ,gMR% re-late the molecular fixed frame to a frame fixed on the rotThese angles are random variables in a powder. The throspace dipolar coupling between two spinsj andk is given by
bjk52m0
4p
g jgk\
r jk3 , ~20!
wherer jk is the spin–spin internuclear distance. The fact
k5343~ i 2e2 i ~p/14!!
768A2p~21!
represents the scaling factor of the homonuclear recoupsequence. The C144
25 sequence with C05(2p)0 has a scal-ing factor uku>0.157. This is slightly higher than that obtained for C72
1 with C05(2p)0(2p)p ~Ref. 14! or C0
5(p/2)0(2p)p(3p/2)0 ,16 which both have a scaling factouku>0.155. The definition of the scaling factor used here,~10! differs from that given in Ref. 16 by a factor of 2/3.
When a recoupling sequence is applied with overalphaseF0 and durationt, starting from time pointt0, andending at time pointt1, the corresponding propagation speroperator is
V~ t1,t0,F0!5exp$2 iHC comm~ t0,F0!t%, ~22!
whereHC commdenotes the commutation superoperator49 of theaverage HamiltonianH (1). We assume that the averagHamiltonian is independent of the time intervalt. This as-sumption is tested later by comparison with accurate simtions.
From Eq.~19! the form of the propagation superoperatis
V~ t1,t0,F0!5Rz~F02 12v r t
0!V0~t!Rz~2F01 12v r t
0!,~23!
whereRz(f) is the rotation superoperator49
Rz~f!5exp$2 ifSzcomm% ~24!
andSzcomm is the superoperator for commutation with the t
tal spin angular momentum along thez axis. V0(t) is the
of
r.gh
g
.
f
a-
propagator for a C14425 sequence with durationt, starting at
time point t50 and an overall phaseF050:
V0~t!5exp$2 iHC comm~0,0!t%. ~25!
Equation~23! indicates an explicit relationship between theffective phase of the propagation superoperator andpulse sequence timings. In the experiments, this link isploited to allow arbitrary incrementation of the evolution iterval t1 .
C. Double-quantum spectra
Consider the pulse sequence in Fig. 7. The~21!-quantum coherence generated by the lastp/2 pulse is de-tected at time pointt2 , using pulse sequences with differeevolution intervalst1 . The complex 2D signal amplitudemay be written
s~ t1 ,t2!5 i ~SzuSz!21(S2uU0~ t2,0!RxS p
2 D3VRU0~ tR
0,tE1 !VEuSz), ~26!
where U0(tb ,ta) expresses free propagation in the abseof rf fields, from time pointta to time pointtb . VE and VR
are shorthand notations forVE5V(tE1,tE
0,FE) and VR
5V(tR1,tR
0,FR). The factori takes into account quadratursignal detection.42 In addition a normalization facto(SzuSz)
21 has been included. The~21!-quantum operatormay be written as a superposition of terms from individuspins:
(S2u5(l
(Sl2u5(
l~Slx2 iSlyu. ~27!
If the experiment is performed far from rotationresonance,1–6 and CSA modulations are ignored, the evoltion of the individual~21!-quantum coherences may be wrten as
~Sl2uU0~ t2,0!>exp$~ iv l
iso2l l !t2%~Sl2u, ~28!
wherev liso is the isotropic chemical shift of spinsSl , andl l
is the coherence decay constant. The result of the finalp/2pulse may be written
(Sl2uRxS p
2 D5~Slx1 iSlzu ~29!
which leads to the following expressions for the 2D signamplitude, neglecting CSA and rotational resonance effe
s~ t1 ,t2!>~SzuSz!21(
l~2 iSlx1SlzuVRU0~ t1!VEuSz!
3exp$~ iv liso2l l !t2%. ~30!
The evolution propagatorU0(tb ,ta) has been written asU0(t1), for the sake of brevity. Since the Hamiltonian in thabsence of an rf field commutes withSz , the central propa-gator may be written as
VRU0~ t1!VE5Rz~FR2 12v r tR
0 !VR0Rz~2FR1FE1 1
2v r~ tR02tE
0 !!U0~ t1!VE0Rz~2FE1 1
2v r tE0 !. ~31!
ae
rin
p
ne
s:
nd
toinsein
ar
as
-
e
-co-
-
ese
f
the
d
m-
rale
vals
8547J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Helical pulse sequences in MAS–NMR
Using the relationship between the pulse sequence phand timings@Eq. ~14!#, and taking into account the phascycling, which selects double-quantum coherences duthe evolution intervalt1 as well asz-magnetization at timepoint tR
1, we get
s~ t1 ,t2!52~SzuSz!21
3(l
~SlzuVR0 P~2!U0~ t1!VE
0 uSz!
3exp$~ iv liso2l l !t2%, ~32!
where the overbar denotes the phase-cycled signal amtude. The termsVE
0 andVR0 are abbreviations forV0(tE) and
V0(tR), respectively. The termP(2) represents a projectiosuperoperator for~62!-quantum coherences, which may bwritten in terms of the individual spin operators as follow
P~2!5 P~12!1 P~22! ~33!
with
P~12!5(j ,k
uSj1Sk
1)~Sj1Sk
1u~Sj
1Sk1uSj
1Sk1!
,
~34!
P~22!5(j ,k
uSj2Sk
2)~Sj2Sk
2u~Sj
2Sk2uSj
2Sk2!
.
The sum is to be taken over all spin pairs a(Sj
1Sk1uSj
1Sk1)5(Sj
1Sk1uSj
1Sk1)52N22. For simplicity, we
have ignored contributions to the projection superoperacorresponding to double-quantum coherences involvmore than two spins. There is no contribution from theterms in the case of two and three spin systems in genand in the case of a multiple-spin system for short timetervalstE andtR .
If rotational resonance effects and CSA modulationsignored, the evolution of the individual~62!-quantum coher-ences has a simple form:
U0~ t1!uSj1Sk
1)>uSj1Sk
1)exp$2 i ~v jiso1vk
iso!t12l jkt1%,~35!
U0~ t1!uSj2Sk
2)>uSj2Sk
2)exp$1 i ~v jiso1vk
iso!t12l jkt1%,
wherel jk is a double-quantum decay rate constant.These expressions may be combined to obtain the ph
cycled 2D spectral amplitudes:
s~ t1 ,t2!>(j ,k
(l
~sjk→ l~12!~ t1 ,t2!1sjk→ l
~22!~ t1 ,t2!!, ~36!
where
sjk→ l~12!~ t1 ,t2!5ajk→ l
~12! exp$2 i ~v jiso1vk
iso!t11 iv lisot2
2l jkt12l l t2%,~37!
sjk→ l~22!~ t1 ,t2!5ajk→ l
~22! exp$1 i ~v jiso1vk
iso!t11 iv lisot2
2l jkt12l l t2%,
and the 2D signal amplitudes~for a single molecular orientation! are
ses
g
li-
rgeral-
e
e-
ajk→ l~12!~tE ,tR!52N(
m~SlzuVR
0 uSj1Sk
1!~Sj1Sk
1uVE0 uSmz!,
~38!
ajk→ l~22!~tE ,tR!52N(
m~SlzuVR
0 uSj2Sk
2!~Sj2Sk
2uVE0 uSmz!.
The normalization factorN is given by
N5~Sj1Sk
1uSj1Sk
1!21~SzuSz!21
5~Sj2Sk
2uSj2Sk
2!21~SzuSz!215~N22N24!21. ~39!
The termajk→ l(12) represents the complex amplitude of th
2D spectral peak at the frequency coordinates (v1 ,v2)5(2v j
iso2vkiso,v l
iso). The termajk→ l(22) represents the com
plex amplitude of the 2D spectral peak at the frequencyordinates (v1 ,v2)5(v j
iso1vkiso,v l
iso). The amplitudes de-pend on the orientationVMR through Eq.~19!. In a powder,the orientational average is observed:
^ajk→ l~62! &VMR
51
8p2 E0
2p
daMRE0
p
dbMR sinbMR
3E0
2p
dgMR ajk→ l~62!~VMR!. ~40!
The 2D spectral peaks fall into two classes:~i! Peaks ofthe form (jk→ j ) which represent transfer of doublequantum coherence between spinsSj and Sk into single-quantum coherence of a spin within the same pair. Thpeaks here referred to asdirect double-quantum peaks.~ii !peaks of the form (jk→ l ), which represent transfer odouble-quantum coherences between spinsSj and Sk intosingle-quantum coherence of a third spinSl . These peaks arereferred to asindirect double-quantum peaks.
D. Spectral amplitudes
The theoretical expressions Eq.~38! may be used to in-vestigate the dependence of the 2D peak amplitudes onexcitation and reconversion intervalstE and tR . As shownin Appendix B, the Liouvillian matrix elements are relatethrough
~SlzuV0~t!uSj1Sk
1!5~SlzuV0~t!uSj2Sk
2!* ,~41!
~Sj1Sk
1uV0~t!uSlz!5~Sj2Sk
2uV0~t!uSlz!* .
The ~62!-quantum spectral amplitudes are therefore coplex conjugates of each other:
ajk→ l~22!5~ajk→ l
~12! !* . ~42!
This relationship is not sufficient to prove that 2D spectpeaks are pure absorption, after cosine transform in tht1
dimension. As discussed in standard texts,39 pure absorptionspectra are obtained only if the amplitudes for the~62!-quantum pathways arereal, as well as being equal.
It is useful to expand the terms in Eq.~38! with respectto the pulse sequence excitation and reconversion intertE andtR , using
-r o-o
e
urrmr
ith
fo
he
2ex
re
al,ab-le-
a-
8548 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
~Sj1Sk
1uV0~t!uSlz!5Tr$Sj2Sk
2(Slz1 i t@H ~1!,Slz#
1~ i t!2
2!@H ~1!,@H ~1!,Slz##1¯)%.
~43!
Some labor is saved by using the symmetry relationship
~SlzuV0~t!uSj6Sk
6!52~Sj6Sk
6uV0~t!uSlz!* ~44!
which applies ifJ couplings are ignored, as proved in Appendix B. If the expansions are carried out to a total orde4 for the tE and tR intervals, we get the following expressions for the direct double-quantum peaks, for a single mlecular orientation:
ajk→ j~12! ~tE ,tR!
>1
N H 1
2uv jku2tEtR
21
12uv jku2S (
lÞ juv j l u21 (
lÞ~ j ,k!uvklu2D tE
3tR
21
48uv jku2S 4(
lÞ juv j l u21 (
lÞ~ j ,k!uvklu2D tEtR
31¯J .
~45!
The sum( lÞ j implies a sum over all values ofl not equal toj, while the sum( lÞ( j ,k) implies a sum over all values oflnot equal to eitherj or k. For the indirect double-quantumpeaks we get for a single molecular orientation
ajk→ l~12!~tE ,tR!
>1
N H 21
16uv jku2~ uv j l u21uvklu2!tEtR
31¯J . ~46!
If the expansions are carried out to total order of 6 intE andtR , very long expressions are obtained, which arenot real ingeneral. Similar series expansions have been given prously for the build-up oftotal multiple-quantum filtered sig-nal amplitudes.50–52
The following conclusions can be drawn.
~i! All 2D amplitudes arereal for small values oftE andtR . Under these conditions, the 2D spectra are in pabsorption, after applying a cosine Fourier transfoin the t1 dimension. Pure absorption 2D spectra anot obtained for single orientations in systems wN.3, if large values oftE or tR are used.
~ii ! The indirect and direct peaks have opposite signs,small values oftE andtR .
~iii ! The indirect peaks vanish for small values of tdouble-quantum reconversion intervaltR .
Systems of two or three spins are special cases. Theamplitudes may be solved analytically, and the resultingpressions are real for all values oftE andtR . For two-spinsystems, only direct double-quantum peaks exist. Theevant Liouvillian matrix elements are
f
-
vi-
e
e
r
D-
l-
~Sj1Sk
1uV0~t!uSjz!5i
2
v jk*
uv jkusin~ uv jkut! ~47!
leading to the double-quantum peak amplitudes
ajk→ j~12! ~tE ,tR!5ajk→ j
~22! ~tE ,tR!
5 14sin~ uv jkutE!sin~ uv jkutR!. ~48!
Since the~62!-quantum peak amplitudes are real and equthe two-dimensional double-quantum spectra are in puresorption, if a cosine Fourier transform is used in the doubquantum dimension.
For three-spin systems, the relevant Liouville space mtrix elements are
~Sj1Sk
1uV0~t!uSjz!52iv jk*
3A3v rms3
sinSA3
2v rmst D
3H uvklu21cosSA3
2v rmst D
3~ uv jku21uv j l u2!J ,
~49!
~Sj1Sk
1uV0~t!uSlz!528iv jk*
3A3v rms3
sin3SA3
4v rmst D
3cosSA3
4v rmst D $uv j l u21uvklu2%,
where the root-mean-square recoupled interactionv rms isgiven by
v rms51
A3~ uv12u21uv13u21uv23u2!1/2. ~50!
The two-dimensional peak amplitudes in Eq.~38! evaluate tothe following expressions:
ajk→ j~12! ~tE ,tR!5ajk→ j
~22! ~tE ,tR!
51uv jku2
27v rms4 sin~)v rmstE!sinSA3
2v rmstRD
3H uvklu21cosSA3
2v rmstRD
3~ uv jku21uv j l u2!J ~51!
and
ajk→ l~12!~tE ,tR!5ajk→ l
~22!~tE ,tR!
524uv jku2
27v rms4
sin~A3v rmstE!
3sin3SA3
4v rmstRD cosSA3
4v rmstRD
3$uv j l u21uvklu2%. ~52!
e
inon
s,d
lerys2o
ona
uito
tio
c
edhea
e
m
ho
dipun
-athu
ing
tsat-eri-
eue
aks
tu
ak
andmed
8549J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Helical pulse sequences in MAS–NMR
Similar results have been presented before for the castE5tR ~Refs. 17 and 53! and in the context of zero53 andtriple-quantum54 excitation. The expressions for two-spand three-spin systems are independent of the orientatianglegMR .
Equations~51! and~52! show that in three-spin systemthe amplitudes for the~62!-quantum pathways are real anequal. This implies that the two-dimensional doubquantum spectra are in pure absorption after cosine Foutransformation in the double-quantum dimension. For stems of more than three spins, one obtains nonabsorptionamplitudes for single molecular orientations in the caselong tE or tR. However, simulations indicate that absorptispectra are obtained after powder averaging even in this c
E. Double-quantum spectra
Experimental results for 98% labeled13C3–L-alanine@Fig. 8~a!# at B054.7 T and a spinning frequency ofv r /2p511.000 kHz are shown in Figs. 9–11. The sample was pchased from Cambridge Isotope Laboratories and used wout further purification. The experiments were performeda Chemagnetics Infinity-200 spectrometer using a filledmm zirconia rotor.
The spectra were obtained using a cross-polarizacontact time of 800ms. C144
25 cycles were used for bothexcitation and reconversion of double-quantum coherenThe excitation part of the sequence consisted ofqE517cycles, which corresponds to an excitation time oftE
5441.5ms. The evolution intervalt1 was incremented insteps of 21.74ms. Continuous wave decoupling was uswith the proton nutation frequency 100 kHz during tC144
25 sequences, and 74 kHz during the evolution andquisition intervals. The signal in thet1 dimension wasapodized with a cos2 function and converted into the timdomain using a cosine Fourier transform.
Figure 9~a! shows the experimental 2D double-quantuspectrum for a reconversion intervaltR5207.8ms. Note theabsence of the indirect double-quantum peaks for this svalue of the reconversion interval. Figure 10~a! shows ex-perimental results for a long reconversion intervaltR
51168.8ms. The spectrum displays strong direct and inrect peaks of both signs. In both cases all peaks are inabsorption phase, as predicted. Note that the negative siga peak doesnot indicate that it is indirect.
Figures 9~b! and 10~b! show numerically exact simulations of the two-dimensional spectra, using the spin intertion parameters given in Ref. 55. The simulations mimicexperiments as closely as possible only assuming rectang
FIG. 8. Molecular structure in the samples used for the double-quanexperiments:~a! L-alanine;~b! L-tyrosine.
of
al
-ier-Df
se.
r-h-n4
n
e.
c-
rt
-reof
c-elar
pulses and neglect of relaxation and were performed usthe COMPUTE algorithm55,56 in the t2 dimension for opti-mal computational efficiency. Agreement with experimenis good, with the exception of the peak amplitudes. Wetribute these discrepancies to relaxation effects in the expmental system~see below!. The simulations in Fig. 10~b!employ an rf field with an amplitude equal to 99.4% of thnominal value. The simulation using the nominal rf valdisplays small phase distortions.
Figure 11 shows the measured integrals of the 2D peas a function of the reconversion intervaltR , with the exci-tation interval remaining fixed at the valuetE5441.5ms.
m
FIG. 9. ~a! Experimental 2D double-quantum13C spectra of 98%13C3–L-alanine, obtained using the pulse sequence in Fig. 7 withtE
5441.5ms and tR5207.8ms. Note the absorption mode spectral peshapes and the absence of indirect double-quantum peaks.~b! Accurate nu-merical simulations using the known geometry of the three-spin systemthe simulation parameters in Ref. 55. Powder averaging was perforusing 1154 orientationsVMR , chosen according to the ZCW scheme~Ref.62!.
FIG. 10. As in Fig. 9, but withtE5441.5ms andtR51168.8ms. Negativepeaks denoted by gray contour lines.
s
of
cal scale
8550 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
FIG. 11. Main plots: Symbols: Experimental integrals of the 9 spectral peaks in the 2D double-quantum13C spectra of 98%13C3–L-alanine, obtained usingthe pulse sequence in Fig. 7 as a function of the reconversion intervaltR . The excitation interval was fixed attE5441.5ms. The plots show the amplitudeof peaks with the following frequency coordinates (v1 ,v2): ~a! (va
iso1vbiso ,vCO
iso); ~b! (vaiso1vb
iso , vaiso); ~c! (va
iso1vbiso , vb
iso); ~d! (vCOiso1vb
iso , vCOiso); ~e!
(vCOiso1vb
iso , vaiso); ~f! (vCO
iso1vbiso , vb
iso); ~g! (vCOiso1va
iso , vCOiso); ~b! (vCO
iso1vaiso , va
iso); ~i! (vCOiso1va
iso , vbiso). Solid lines: accurate numerical simulations
the amplitudes using the known geometry andJ couplings and including damping. Dashed lines: Analytical solutions given in Eqs.~51! and ~52! includingdamping. In both cases the same powder averaging as in Fig. 9~b! was performed. The phenomenological relaxation time constants used in Eq.~54! are givenby ~a!–~c!; Tab,CO51.34 ms,Tab,a52.34 ms,Tab,b52.28 ms; ~d!–~f!, TaCO,CO526.1 ms,TaCO,a51.57 ms,TaCO,b51.58 ms; ~g!–~i!, TCOb,CO51.18 ms,TCOb,a51.05 ms,TCOb,a51.22 ms. Insets: solid lines: simulations using damping. Dotted–dashed lines: simulations without damping. Note the vertiof ~d!–~f!.
lyen
attetho
n-seu-a
uliv
-ic
ofuslyex-
se
ac-ise
cal-nts
cale-themp-
o-
is
ns,od.
olv-ri-ns,es.
The slow initial build up of the indirect peaks is visualobvious. Effects involving all three spins are already evidfor very short reconversion intervals.
The solid lines in the figure are the results of accurthree-spin simulations, using the spin interaction paramegiven in Ref. 55. These simulations use a calculation oftwo-dimensional double-quantum amplitudes according t
ajk→ l~62!8~tE ,tR!52N(
m~SlzuU~ tR
1,tR0 !uSj
6Sk6!8
3~Sj6Sk
6uU~ tE1,tE
0 !uSmz!8, ~53!
where U(t1,t0) is the numerical calculated propagator, icluding all spin interactions and the evolution interval isto zero (tE
15tR0). In practice the matrix elements were calc
lated in Hilbert space. This equation corresponds to theerage Hamiltonian equation@Eq. ~38!#, but incorporating aphenomenological correction for relaxation during the mtiple pulse sequence. The modified matrix elements are gby
~Sj6Sk
6uU~ tE1,tE
0 !uSmz!85~Sj6Sk
6uU~ tE1,tE
0 !uSmz!e2t/Tjk,m,
~54!
~SlzuU~ tR1,tR
0 !uSj6Sk
6!85~SlzuU~ tR1,tR
0 !uSj6Sk
6!e2t/Tjk,l.
Each of the nine independent time constantsTjk,l withj ,k,l 5CO,a,b denoting the corresponding13C site, corre-sponds to the transfer of an initialz-magnetization component to an individual double-quantum coherence. In pract
t
erse
t
v-
-en
e,
the nine time constants may be divided into three groupsthree. Each group of three may be estimated simultaneoby a least square fit to three experimental curves. Forample, Tab,CO, Tab,a , and Tab,b are estimated by fittingaab→CO
(62) 8, aab→a(62) 8, andaab→b
(62) 8 to the experimental resultin Figs. 11~a!–11~c!. Similar estimations can be done for thtwo remaining groups of damping time constants. In prtice, the quality of the fit was not very sensitive to the precvalue of these relaxation time constants.
The dashed lines in Fig. 11 are given by the analytiformula Eq.~51! and Eq.~52!, but also use the phenomenological damping factors. The same damping time constaare used as for the solid lines.
The inset plots in Fig. 11 compare the exact numerisimulations with and without damping. The solid line corrsponds to the damped curves in the main plot, whiledotted–dashed line represents the simulations without daing.
We have compared these simulations with full twdimensional spectral simulations shown in Figs. 9~b! and10~b!. The difference between the two simulation resultsinsignificant.
The agreement between the exact numerical simulatioanalytical results, and the experimental amplitudes is goThe deviations are largest for double-quantum peaks inving CO, which is due to its large CSA. The plotted expemental amplitudes only contain the centerband contributioand slightly underestimate the full magnetization amplitud
um
s
b-
bd
fuhms
-re
n
ee
ae-rothias
yh
atiohs
by
na-ightheeri-of
bedr ofnsbed
eedal-
l be
ay.
ombleym-of
ee-Mbeop-ent
iledleds.
ugesesd.mseas-lyen-wn
m-
edi-hasrantnkand
ythhe
m
8551J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Helical pulse sequences in MAS–NMR
Figure 12 shows an experimental 2D double-quantspectrum of @98% – U–13C#–L-tyrosine @Fig. 8~b!# at B0
59.4 T and a spinning frequency ofv r /2p520.000 kHz.In this case the supercycled sequence SC14 was uThe excitation part of the sequence consisted ofqE528cycles, i.e., the excitation sequence was givenC144
5•@P0
21•C144
25•P0] p/7 . The reconversion part of the se
quence consisted ofqR57, i.e., seven elements of@C1445#p .
The excitation and reconversion time intervals were giventE5400.1ms andtR5100.0ms. The sample was purchasefrom Cambridge Isotope Laboratories and used withoutther purification. The experiments were performed on a Cmagnetics Infinity-400 spectrometer using a filled 3.2 mzirconia rotor. The spectra were obtained using a cropolarization contact time of 2.5 ms. The evolution intervalt1
was incremented in steps of 10ms. Continuous wave decoupling was used during SC14 with a proton nutation fquency of 227 kHz. TPPM decoupling57 was used during theevolution intervalt1 and the acquisition. The proton nutatiofrequency was 96 kHz. The signal in thet1 dimension wasapodized with a cos2 function and converted into the timdomain using a cosine transform. The experimental 2D sptrum permits a straightforward assignment of the13C peaks,as in the liquid state 2D-INADEQUATE experiment.32,33 Aslight nonequivalence of thed, d8 and e, e8 sites may bediscerned.
IV. CONCLUSIONS
The results given in this paper may be summarizedfollows: ~i! It is possible to develop a general theory of rcoupling in magic-angle-spinning solids, exploiting synchnized helical modulations of the space and spin parts ofnuclear magnetic interactions. The average Hamiltonproperties of such sequences may be formulated in termthe three symmetry numbers for the pulse sequence.~ii !These principles may be applied to the problem of spin stem recoupling at high spinning frequencies. A sequence
FIG. 12. Experimental 2D double-quantum13C spectrum of@98% – U–13C# –L-tyrosine, at a field ofB059.4 T and a spinning frequencof v r /2p520.000 kHz, obtained using the pulse sequence in Fig. 7 withsupercycle SC14 for excitation and reconversion of double-quantum coence. The excitation and reconversion intervals weretE5400.1ms andtR
5100.0ms. The assignments of the double-quantum coherences to thelecular site pairs are indicated along the right-hand axis.
ed.
y
y
r-e-
s-
-
c-
s
-enof
s-as
been demonstrated which employs an exceptional low rof rf field to spinning frequency while still being robust witrespect to rf amplitude errors.~iii ! This class of sequencemay be compensated with respect to chemical shift effectsemploying a supercycle scheme.~iv! The dynamics of re-coupled multiple-spin systems have been thoroughly alyzed in the context of double-quantum spectroscopy at hMAS frequencies. Good agreement is obtained betweenexperimental spectra and analytical results as well as numcally exact simulations, in the case of a three-spin systemknown geometry.
The principles of pulse sequence construction descrihere are very general and may be applied to a numbeother problems in magic-angle-spinning NMR. Applicatioto heteronuclear decoupling have been descrielsewhere.35 In addition, the symmetries C31
0 and C410 have
been used for selection of isotropic Hamiltonians58 andhomonuclear decoupling,59 respectively. These principles arimmediately predictable from the symmetry rules describin the current paper. Generalized symmetry rules whichlow the treatment of heteronuclear pulse sequences wildescribed elsewhere.
The C1445 sequence described in this paper is not
unique solution for recoupling at high spinning frequencAs indicated in Table I, a very large number of useful CNn
n
symmetries exist, and each symmetry permits great freedin the choice of the elemental cycle. In addition, it is possito construct sequences with the same interaction frame smetry as CNn
n sequences, but with elements consistingz-rotations instead of cycles. This additional degree of frdom has been discussed in the context of TPPdecoupling57 in Ref. 35. The results presented here shouldregarded as defining a framework guiding the search fortimal sequences. In the end, the choice of the basic elemand the type of symmetry must be guided by more detaconsiderations such as the particular form of the recoupHamiltonian, and the properties of the higher-order termThe solutions discussed in this paper is only one of a hnumber of possibilities. Furthermore, new symmetry claswith different selection rules have recently been describe40
Two-dimensional spectroscopy of multiple-spin systeat high spinning frequencies is expected to become incringly important, especially in the solid-state NMR of heavilabeled biomolecules. The good agreement of the experimtal three-spin dynamics with the theoretical curves, as shoin Fig. 11, bodes well for the precise control of spin dynaics in such experiments.
ACKNOWLEDGMENTS
The research was supported by the Go¨ran GustafsonFoundation for Research in the Natural Sciences and Mcine, and the Swedish Natural Science Foundation. A.B.been supported by the Marie Curie Research Training GERBFMBICT961439 from the European Union. We thaO. G. Johannessen for experimental help and A. SebaldS. Dusold for discussions.
er-
o-
ro
ioi
eeint.ve
ex
se
in
r
oss-een
are
e-
rfs, in-
andefore
of
o
e--
el
the
8552 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
APPENDIX A: SUPERCYCLED C Nnn SEQUENCES
The spin Hamiltonian in the presence of a rotor synchnized rf pulse sequence is given by
H~ t !5H int~ t !1H rf~ t !, ~A1!
where the internal spin HamiltonianH int(t) is time depen-dent because of the sample rotation, while the rf spin HamtonianH rf(t) is time dependent because of the modulationthe rf fields. It is convenient to analyze the spin dynamicsthe interaction frame of the rf field, i.e.,
H~ t !5W~ t !†H int~ t !W~ t !, ~A2!
where the frame transformation operatorW(t) solves theequations
d
dtW~ t !52 iH rf ~ t !W~ t !, ~A3!
W~ t0!51. ~A4!
These equations maintain a notational distinction betwoperators which connect states at two different time po~‘‘propagators’’! and those that depend on one time poin60
The propagator under the interaction frame Hamiltonian othe durationT of the pulse sequence~starting at time pointt0! is expressed in terms of an effective Hamiltonian
U~T1t0,t0!5exp$2 iH T%, ~A5!
whereH is usually approximated as a truncated Magnuspansion
H>H ~1!1H ~2! ~A6!
with
H ~1!5T21Et0
t01Tdt H~ t !, ~A7!
H ~2!5~2iT !21Et0
t01Tdt8E
t0
t8dt@H~ t8!,H~ t !#. ~A8!
The internal part of the spin Hamiltonian may be expresusing terms with different spherical rankl for rotations ofthe spin polarizations
H int~ t !5(l
Hl~ t !, ~A9!
whereHl(t) is proportional to the irreducible spherical spoperatorTl0 . For example, chemical shift terms~both iso-tropic and anisotropic! have l51, while homonucleardipole–dipole coupling terms havel52. The Magnus ex-pansion terms may then be written
H ~1!5(l
Hl~1! , ~A10!
H ~2!5 (l2 ,l1
Hl23l1
~2! , ~A11!
where
Hl~1!5T21E
t0
t01Tdt Hl~ t !, ~A12!
-
l-fn
nts
r
-
d
Hl23l1
~2! 5~2iT !21Et0
t01Tdt8E
t0
t8dt@Hl2
~ t8!,Hl1~ t !#,
~A13!
and
Hl~ t !5W~ t !†Hl~ t !W~ t !. ~A14!
The symmetry rules given in Eq.~8! may be used to engineesequences with desirable properties forH (1). For example,the chemical shift anisotropy components ofH (1) vanish forthe symmetries C72
1, C1445, etc., as given in Table I. It is also
possible to predict the destruction of a large number of crterms of the formH231
(2) .35 However, the remaining highorder terms cause an undesirable interference betwdipole–dipole recoupling and chemical shift anisotropy.
In order to see how to eliminate such terms, compnow two sequences with opposite spin winding numbern,i.e., CNn
n and CNn2n . For brevity, these sequences are d
noted (C1) and (C2). The rf spin Hamiltonians for thesetwo sequences are related through
H rf~C1)5Rx~p!H rf~C
2)Rx~p!†, ~A15!
whereRx(p)5exp$2ipSx%. This equation assumes that allphases are reversed in sign between the two sequencecluding any internal phases within the C elements~this lattercondition is unnecessary if the C elements only contain 0p phase shifts!. The frame transformation operators for thtwo sequences at corresponding points of time are thererelated through
W~C1;t !5Rx~p!W~C2;t !Rx~p!†. ~A16!
The interaction frame spin Hamiltonians for an interactionspin rankl are related by
Hl~C1;t !5W~C1;t !†Hl~ t !W~C1;t !
5Rx~p!W~C2;t !†Rx~p!†Hl~ t !
3Rx~p!W~C2;t !Rx~p!†
5~21!lRx~p!Hl~C2;t !Rx~p!†. ~A17!
The last line follows from the transformation property
Rx~p!Tl0Rx~p!†5~21!lTl0 . ~A18!
It follows that the average Hamiltonian terms for the twsequences are related by
Hl~1!~C1!5~21!lRx~p!Hl
~1!~C2!Rx~p!†, ~A19!
Hl23l1
~2! ~C1!5~21!l21l1Rx~p!Hl23l1
~2! ~C2!Rx~p!†.
~A20!
Equation~A19! shows for example, that two CNnn sequences
with opposite signs ofn cannot be concatenated without dstroying the desirableg-encoding of the double-quantum recoupling.
Now consider a hypothetical situation in which thCNn
2n sequence is bracketed by two infinitely strong ideappulses of phase 0. Denote this bracketed sequence by
’’Th
de-lr
tio
sed
isca.rd
ol
ed
i-e
he
edo
tinge
ral
t-
Eq.
q.
8553J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Helical pulse sequences in MAS–NMR
symbol (PC2P). The effect of thep pulse is to rotate theoverall propagator by the anglep around thex axis, whichleads to the relationships:
Hl~1!~C1!5~21!lHl
~1!~PC2P!, ~A21!
Hl23l1
~2! ~C1!5~21!l21l1Hl23l1
~2! ~PC2P!. ~A22!
Now suppose that the CNnn sequence and the ‘‘bracketed
CNn2n sequence are concatenated to build a supercycle.
idealized supercycle is denoted (C1PC2P). The first orderaverage Hamiltonian term for the supercycle is given by
Hl~1!~C1PC2P!5 1
2Hl~1!~C1!1 1
2Hl~1!~PC2P!. ~A23!
This has the property
Hl~1!~C1PC2P!5H 0 for l5odd,
Hl~1!~C1! for l5even
. ~A24!
It follows that this type of supercycle preserves the first oraverage Hamiltonian forl52 terms ~such as dipolar couplings!, while destroying l51 terms ~such as chemicashifts!. These are desirable properties for homonuclearcoupling sequences.
For the second-order term, the corresponding calculagives
Hl23l1
~2! ~C1PC2P!5 12 Hl23l1
~2! ~C1!
2i
4@Hl2
~1!~PC2P!,Hl1
~1!~C1!#
1 12 Hl23l1
~2! ~PC2P!. ~A25!
For most of the interesting cross terms, the individualquences are designed so as to eliminate the corresponHl
(1) terms, so that the commutator may be neglected. Thtrue, for example, for the cross terms involving the chemishift anisotropy, in the case of C72
1 and C1445 sequences
Under these conditions, the theorem for the second ocross terms is
Hl23l1
~2! ~C1PC2P!
5H 0 for l21l15odd,
Hl23l1
~2! ~C1! for l21l15even.~A26!
This theorem shows that the second order cross term inving the chemical shift anisotropy~l51! and homonucleardipole–dipole coupling~l52! vanishes for the (C1PC2P)supercycle.
In practice, it is not possible to implement the idealizshortp pulses required by the (C1PC2P) supercycle. How-ever cyclic permutationof a p pulse achieves an approxmately similar effect.61 A small additional phase shift of thpulse sequence is used to compensate for the finite timeterval spanned by the permuted element. The symmetry trems given in Eqs.~A24! and ~A26! apply only approxi-mately to supercycles employing cyclically permutelements, but accurate simulations indicate that the apprmation is reasonable.
is
r
e-
n
-ingisl
er
v-
in-o-
xi-
The residual error terms are reduced further by repeathe entire sequence with ap phase shift, which generates thSC14 supercycle written in Eq.~12!.
APPENDIX B: SYMMETRIES OF DOUBLE-QUANTUMSPECTRAL AMPLITUDES
In this appendix, we prove Eqs.~41! and~44!, which areimportant for the calculation of double-quantum spectpeak amplitudes.
First consider the properties of the superoperatorV,given by
V5exp$2 iHC commt%. ~B1!
Here HC comm is the commutation superoperator of the firsorder average Hamiltonian
HC commuA)5u@H ~1!,A#). ~B2!
The superoperatorV is a ‘‘sandwich superoperator,’’49 withthe property
VuA)5uVAV†), ~B3!
where the propagatorV is given by
V5exp$2 i tH ~1!%. ~B4!
The following reasoning may be made:
~AuVuB!* 5~AuVBV†!*
5Tr$~A†VBV†!†%
5Tr$VB†V†A%
5Tr$AVB†V†%5~A†uVuB†!. ~B5!
Equation~41! follows by applying Eq.~B5!, noting thatSlz†
5Slz and (Sj1)†5Sj
2 .In order to prove Eq.~44!, suppose that an operatorR
exists with the following properties:
RR†51,
RVR†5V†,~B6!
RAR†5A,
RBR†52B.
The following reasoning may be developed:
~AuVuB!* 5~A†uVuB†!
5Tr$AVB†V†%
5Tr$RAR†RVR†RBR†RV†R†%
52Tr$AV†BV%
52Tr$BVAV†%52~B†uVuA!. ~B7!
This equation establishes the symmetry relationship in~44!, if the following identifications are madeA5Slz , B5Sj
6Sk6 .
It remains to identify an operatorR with the properties inEq. ~B6!. If the average Hamiltonian has the form of E
ys
m
ett
m
m
et
. M
. M
Theone
e,
.
. H.
r-L,
ys.
.
8554 J. Chem. Phys., Vol. 112, No. 19, 15 May 2000 Brinkmann, Eden, and Levitt
~17!, and theJ couplings are ignored in Eq.~18!, then theoperator
R5expH 2 ip
2SzJ ~B8!
has the appropriate properties.
1D. P. Raleigh, M. H. Levitt, and R. G. Griffin, Chem. Phys. Lett.146, 71~1988!.
2Z.-H. Gan and D. M. Grant, Mol. Phys.67, 1419~1989!.3M. H. Levitt, D. P. Raleigh, F. Creuzet, and R. G. Griffin, J. Chem. Ph90, 6347~1990!.
4T. Karlsson and M. Levitt, J. Chem. Phys.109, 5493~1998!.5M. Helmle et al., J. Magn. Reson.140, 379 ~1999!.6T. Karlsson, M. Ede´n, H. Luthman, and M. H. Levitt, J. Magn. Reson.~inpress!.
7K. Nomura, K. Takegoshi, T. Terao, K. Uchida, and M. Kainosho, J. AChem. Soc.121, 4064~1999!.
8T. Gullion and J. Schaefer, Adv. Magn. Reson.13, 57 ~1989!.9R. Tycko and G. Dabbagh, J. Am. Chem. Soc.113, 9444~1991!.
10D. M. Gregoryet al., Chem. Phys. Lett.246, 654 ~1995!.11M. Baldus, M. Tomaselli, B. Meier, and R. R. Ernst, Chem. Phys. L
230, 329 ~1994!.12M. Baldus and B. H. Meier, J. Magn. Reson.128, 172 ~1997!.13N. C. Nielsen, H. Bildsøe, H. J. Jakobsen, and M. H. Levitt, J. Che
Phys.101, 1805~1994!.14Y. K. Lee et al., Chem. Phys. Lett.242, 304 ~1995!.15C. M. Rienstraet al., J. Am. Chem. Soc.120, 10602~1998!.16M. Hohwy, H. J. Jakobsen, M. Ede´n, M. H. Levitt, and N. C. Nielsen, J.
Chem. Phys.108, 2686~1998!.17M. Hohwy, C. M. Rienstra, C. P. Jaroniec, and R. G. Griffin, J. Che
Phys.110, 7983~1999!.18Y. Ishii, T. Terao, and M. Kainosho, Chem. Phys. Lett.265, 133 ~1996!.19X. Fenget al., Chem. Phys. Lett.257, 314 ~1996!.20X. Fenget al., J. Am. Chem. Soc.119, 12006~1997!.21M. Hong, J. D. Gross, and R. G. Griffin, J. Phys. Chem. B101, 5869
~1997!.22P. R. Costa, J. D. Gross, M. Hong, and R. G. Griffin, Chem. Phys. L
280, 95 ~1997!.23T. Fujiwara, T. Shimomura, and H. Akutsu, J. Magn. Reson.124, 147
~1997!.24Y. Ishii et al., Solid State Nucl. Magn. Reson.11, 169 ~1998!.25X. Fenget al., J. Am. Chem. Soc.119, 6853~1997!.26A. E. Bennett, J. H. Ok, and R. G. Griffin, J. Chem. Phys.96, 8624
~1992!.27W. A. Dollaseet al., J. Am. Chem. Soc.119, 3807~1997!.28T. S. Balaban, A. R. Holzwarth, K. Schaffner, G.-J. Boender, and H. J
de Groot, Macromolecules31, 7404~1995!.29T. A. Egorova-Zachernyuk, B. van Rossum, G.-J. Boender, and H. J
de Groot, Biochemistry36, 7513~1997!.30L. Braunschweiler and R. R. Ernst, J. Magn. Reson.53, 521 ~1983!.31A. Bax, R. Freeman, and S. P. Kempsell, J. Am. Chem. Soc.102, 4849
~1980!.
.
.
.
.
.
t.
.
.
32A. Bax, R. Freeman, and T. Frenkiel, J. Am. Chem. Soc.103, 2102~1981!.
33A. Bax, R. Freeman, T. Frenkiel, and M. H. Levitt, J. Magn. Reson.43,478 ~1981!.
34M. M. Maricq and J. S. Waugh, J. Chem. Phys.70, 3300~1979!.35M. Eden and M. H. Levitt, J. Chem. Phys.111, 1511~1999!.36W. Magnus, Commun. Pure Appl. Math.7, 649 ~1954!.37We employ an indexing of the Magnus expansion starting at one.
older literature on average Hamiltonian theory uses indices which areless than those given here.
38A. Brinkmann, M. Ede´n, X. Feng, H. Luthman, L. Eriksson, A. Gra¨slund,O. N. Antzutkin, and M. H. Levitt, 39th Experimental NMR ConferencAsilomar, CA, 1998.
39R. R. Ernst, G. Bodenhausen, and A. Wokaun,Principles of NuclearMagnetic Resonance in One and Two Dimensions, 5th ed. ~Clarendon,Oxford, 1988!.
40M. Carravetta, M. Ede´n, X. Zhao, A. Brinkmann, and M. H. Levitt, ChemPhys. Lett.~in press!.
41T. Karlsson, A. Brinkmann, P. J. E. Verdegem, J. Lugtenburg, and MLevitt, Solid State Nucl. Magn. Reson.14, 43 ~1998!.
42M. H. Levitt, J. Magn. Reson.126, 164 ~1997!.43M. H. Levitt and O. G. Johannessen, J. Magn. Reson.142, 190 ~2000!.44A. McDermott, T. Polenova, G. Montelione, K. Zilm, E. Paulsen, R. Ma
tin, and G. Coker, 40th Experimental NMR Conference, Orlando, F1999.
45G. Metz, X. Wu, and S. O. Smith, J. Magn. Reson., Ser. A110, 219~1994!.
46A. Brinkmann, M. Ede´n, and M. H. Levitt~unpublished!.47M. Hong, J. Magn. Reson.136, 86 ~1999!.48D. A. Varshalovich, A. N. Moskalev, and V. K. Khersonskii,Quantum
Theory of Angular Momentum~World Scientific, Singapore, 1988!.49J. Jeener, Adv. Magn. Reson.10, 1 ~1982!.50Y. Yen and A. Pines, J. Chem. Phys.78, 3579~1983!.51J. Baum, M. Munowich, A. N. Garroway, and A. Pines, J. Chem. Ph
83, 2015~1985!.52Y. Ba and W. S. Veeman, Isr. J. Chem.32, 173 ~1992!.53A. K. Roy and K. K. Gleason, J. Magn. Reson., Ser. A120, 139 ~1996!.54M. Eden and M. H. Levitt, Chem. Phys. Lett.293, 173 ~1998!.55M. H. Levitt and M. Ede´n, Mol. Phys.95, 879 ~1998!. Table I of this
paper contains two misprints: The signs ofaPM(Cb) and b(Ca2Cb)should be reversed.
56M. Eden, Y. K. Lee, and M. H. Levitt, J. Magn. Reson., Ser. A120, 56~1996!.
57A. E. Bennett, C. M. Rienstra, M. Auger, K. V. Lakshmi, and R. GGriffin, J. Chem. Phys.103, 6951~1995!.
58M. Baldus and B. H. Meier, J. Magn. Reson.121, 65 ~1996!.59D. E. Demco, S. Hafner, and H. W. Spiess, J. Magn. Reson., Ser. A116,
36 ~1995!.60J. Jeener, Bull. Magn. Reson.16, 35 ~1994!.61M. H. Levitt, R. Freeman, and T. Frenkiel, Adv. Magn. Reson.11, 47
~1983!.62V. B. Cheng, H. H. Suzukawa, and M. Wolfsberg, J. Chem. Phys.59,
3992 ~1973!.